Benedict Evans

1. Key Themes

AI Is As Big As The Internet — But Only As Big As The Internet

Evans' most provocative framing is that AI is a massive transformation, but not uniquely unprecedented. The useful comparison is 1997 internet: exciting, mostly not working yet, and the killer applications haven't been built. The people who "get it" mistakenly assume everyone is already there.

"My most controversial opinion is that I think that AI is as big a deal as the internet or mobile, and only as big a deal as the internet or mobile... we're in 1997. It's very exciting. Most stuff kind of doesn't work yet. Most of the stuff that people are going to do hasn't been built yet." — Benedict Evans 00:00:00

Foundation Model Companies Are Likely Commodity Utilities, Not Platform Winners

Evans draws a sharp distinction between where AI sits in the value stack and where the money will ultimately flow. His core thesis: models lack network effects, face commoditization pressure, and likely end up resembling telecom infrastructure — critical but low-margin.

"Global mobile industry has revenue of about a trillion dollars a year... And the stocks have gone nowhere in 25 years because it's an ex-growth, low margin commodity utility... This was that kind of pivotal moment where the telcos thought that they would do all the stuff that you did on your iPhone. And not only do they not do it, but Apple doesn't do it either. It's all further up stack." — Benedict Evans 00:33:31

"The models don't seem to have network effects. So there doesn't seem to be a winner takes all effect where one of these will run away ahead of the other. So you should have competition indefinitely... then why would you have pricing power?" — Benedict Evans 00:38:28

Distribution Is Becoming the Dominant Moat As Models Commoditize

When the underlying product is largely undifferentiated, the competitive battle shifts entirely to distribution. Evans sees Google, Meta, and Apple using their reach to spray "adequate" AI on every surface — and that may be enough.

"Distribution of an adequate product when the field is basically commodity distribution on brand become a big deal... if you look at survey data on which LLMs people use, even before the Llama thing, Meta was up there between ChatGPT and Gemini. Which, if you're in tech, people have completely written it off. But it was, like, they sprayed it on every surface. It wasn't that bad. It was fine." — Benedict Evans 00:45:31

2. Contrarian Perspectives

The "Jobpocalypse" Predictions Are Made By People Who Don't Understand How Enterprises Actually Work

The commonly held view that AI will rapidly eliminate millions of jobs ignores the reality of enterprise software adoption cycles. The doomers assume businesses can and will move overnight.

"You talk to these doomers on Twitter and they would act like every big company is going to buy ChatGPT tomorrow, and then in two weeks time, they'll fire all their stuff. And these people are morons... Enterprise software sales cycle is like 18 months if you're lucky... I know people aren't going to tear out SAP and replace it with X, Y, Z. Maybe in five, in like three, five, 10 years." — Benedict Evans 00:21:00

You Cannot Predict Which Jobs AI Will Eliminate — The Obvious Ones Are Often Wrong

The conventional approach of scoring jobs by "percentage automatable" is fundamentally flawed. The professions you'd least expect to be disrupted are often the ones that get transformed, while supposedly vulnerable ones survive because the task wasn't actually the job.

"I can't look at a senior partner at a law firm and say, well, 17% of their work could be automated. This is horseshit... The stuff that you don't think is, you can't predict which things are going to be exposed necessarily... things that won't be affected by AI — personal trainers. Okay. So I take my iPhone and I balance it on the metal piece with the camera pointed at me and I ask an AI to build me a training routine and watch me and tell me if I'm doing it right. Why do I need a personal trainer?" — Benedict Evans 00:01:23 / 01:04:23

Sam Altman's "Selling Intelligence Like Electricity" Vision Is Naive About Utility Economics

A widely admired vision — AI as metered infrastructure — actually describes a terrible business model. Evans uses the telecom industry as proof that being indispensable infrastructure doesn't translate to value capture.

"Sam Altman said we're going to be selling AI intelligence on a meter like water or electricity. And you look at this and think, my dear sweet child, you need me to explain the marginal structure of the utility industry to you. Because guess what? When you watch television, the TV company isn't paying a percentage of your monthly bill to the electricity company." — Benedict Evans 00:32:38

Dario Altman's Labor Market Predictions Should Be Discounted — He's Not an Economist

Evans makes a blunt argument against treating AI lab founders as authoritative voices on labor economics, while being careful not to impugn their motives.

"I would place... I don't like argument from authority. If you're going to use argument from authority, then it should be relevant to the field. So I'm interested in Dario's opinions on where models are going to go in the next six to 12 months. Not particularly interested in his opinions on theory of labor and market value and comparative advantage. Like, yeah, maybe he had a course on that at university. So did I." — Benedict Evans 00:18:09

AGI and "Superintelligence" Are Moving Definitional Targets, Not Scientific Milestones

The terms are being retroactively redefined to match whatever current systems can do, making all AGI timeline debates essentially semantic rather than substantive.

"AI is whatever machines can't do yet. Because once machines can do it, people say, well, that's just software... Is superintelligence more than AGI or less than AGI? Because last year I thought superintelligence was really good, but not actual AGI. And now it's like, oh, no, no, we've already got AGI, but superintelligence, that's really hard. It's like all these terms." — Benedict Evans 00:27:23 / 00:28:21

3. Companies Identified

WorkOS

B2B SaaS infrastructure company that provides enterprise authentication features (SSO, SCIM, RBAC, audit logs) as drop-in APIs. Described as "Stripe for enterprise features." Mentioned as the infrastructure powering OpenAI, Anthropic, Cursor, Vercel, Replit, Sierra, and Clay.

"What do OpenAI, Anthropic, Cursor, Vercel, Replit, Sierra, Clay, and hundreds of other winning companies all have in common? They are all powered by WorkOS." — Lenny Rachitsky 00:05:15

Frame.io

Video collaboration platform. Cited as an example of a product that unlocks enormous value not because of new underlying technology, but because someone finally recognized the problem and applied existing web capabilities to it — illustrating how AI applications may emerge.

"A company called Frame.io, video editing, video collaboration. And there's nothing new there that you couldn't have done at least five years earlier and maybe 10 years earlier... The delay was somebody realizing, oh, we could — that problem exists inside that industry. And this is the way that we would solve it." — Benedict Evans 00:22:23

Spotify

Used as the canonical example of what it looks like when a technology unlocks a genuinely new product category rather than just doing the old thing more efficiently. Not an online music store — something categorically different.

"Spotify is not an online music store. It's something else... you have to get past 'we do the old stuff, but more.' And you have to get to what do you do that's different because of this. What is this change? What wasn't possible before?" — Benedict Evans 00:01:28 / 01:01:57

4. People Identified

Independent technology analyst, former A16Z partner and equity research analyst. Publishes a widely-read weekly newsletter and bi-annual presentations on major tech trends. Currently focused on AI's systemic effects.

"Benedict was a longtime partner at A16Z as their in-house analyst and resident thinker. Before that, he was a longtime equity researcher. And for the past six years, he's been an independent analyst tracking the most important tech trends." — Lenny Rachitsky 00:01:14

Steven Sinofsky

Former head of Windows at Microsoft, later partner at A16Z. Cited for a sharp strategic observation about how incumbents respond to platform shifts.

"Steven Sinofsky at A16Z, who used to run Windows, would always say, incumbents always try and make the new thing a feature. And sometimes they're right. Sometimes it's a feature." — Benedict Evans 00:42:51

Dan Shipper

Founder/CEO of Every (AI-focused media/product company). Cited as evidence that even the most AI-forward organizations are growing headcount, contradicting automation-eliminates-jobs narratives.

"I just had Dan Shipper from Every on the podcast. Everyone's just increasing headcount. Like the companies you would think would be least likely to add humans are adding many, many humans." — Lenny Rachitsky 00:17:44

5. Operating Insights

The "Task vs. Job" Framework Is the Right Lens for Evaluating AI's Impact on Your Business

Before automating anything, operators need to honestly separate what the task is from what the job actually is. Automating the task sometimes kills the business model; other times the task was never the point. The elevator attendant is the cautionary tale: the job literally was the task. But McKinsey's job is not making PowerPoints.

"What's the hard part of the job? Is the hard part of the job writing the code line by line?... Or is the hard part of the job something else? Is it the task or the job?... What you actually pay Bain to do is go and walk all over your company and work out, yes, but why is it that you didn't do that? And how do the politics of this work?... The PowerPoint is just like the task. But that's not what you hired them for." — Benedict Evans 00:12:59 / 00:16:06

Building AI Into Enterprise Workflows Requires Dedicated Project Capacity — Which Almost No Company Has

Companies that want to capture AI productivity gains will need to temporarily staff up, not cut, because the workflow redesign work itself requires people. This is the actual reason AI labs are investing in consulting arms.

"You're supposed to completely reimagine all of the internal workflows of your company and work out which of them could be automated really quickly with AI. That's a project. That's a project that needs like five or ten people to sit down and spend a month or two working it out... Well, guess what? Who's going to do that? Because you don't have a bunch of people sitting around not doing anything." — Benedict Evans 00:11:19

For Consumer AI, Marginal Cost Is the Real Barrier to Breakout Apps — Not Product Quality

The reason there are no breakout consumer AI apps yet isn't a product problem — it's a unit economics problem. The playbook of "make it free, get 50M users, monetize later" breaks when every user interaction has meaningful compute cost.

"We don't have breakout consumer AI apps yet because I think because of marginal cost more than anything else, you can't make it free and get 50 million users and then have a revenue model." — Benedict Evans 00:15:23

6. Overlooked Insights

The UK Post Office/Fujitsu Scandal Is a Preview of AI's Institutional Liability Risk

Evans briefly mentioned this story almost as a sidebar, but it's highly significant for anyone building or deploying AI in institutional contexts. A buggy system was used as courtroom evidence. People went to prison. Suicides occurred. The institutions — Post Office and Fujitsu — actively denied bugs existed under oath. As AI systems get embedded into consequential decisions (HR, legal, financial), the liability and institutional cover-up risk this story illustrates is a massive, underpriced risk.

"The post office rolled out this new computer system built by Fujitsu that had a bunch of bugs in it that showed shortfalls in cash. And the post office looks at this and says, aha, we knew these people were stealing from us. Hundreds of people get prison. Bunch of suicides. Bunch of bankruptcies. People lose their homes. Meanwhile, people from the post office and people from Fujitsu are going to court and swearing there's no bugs in the system... every wave of technology comes with ways that you can ruin people's lives either deliberately or by accident." — Benedict Evans 00:57:39

The Absence of Any Public DAU Data From AI Labs Is Itself a Major Signal

Evans dropped this observation almost in passing while discussing anti-AI sentiment and employment data, but it's remarkable: we are in the middle of what is supposed to be the most transformative technology in history, and the companies at the center of it publish zero meaningful usage metrics. This opacity makes it nearly impossible to size the actual market, validate AI productivity claims, or hold anyone accountable for outcomes.

"The model labs don't tell us anything. They don't give us any meaningful usage information. They give us these weird studies of how many people use this for this and that. They don't give us a daily active use number. We do not have a daily active user number for ChatGPT. It's crazy. And all the data comes from academic economists trying to back stuff out of BLS surveys." — Benedict Evans 00:50:30